Results (
Indonesian) 1:
[Copy]Copied!
Wecompared two algorithms: items sharing most attribute values of the purchased items(“close” recommendation condition) and items belonging to the most deeply visitedgeneral category (“broad” recommendation condition). The operationalization of thisvariable was made in the following manner. In the baseline condition (nopersonalization), items were randomly selected in the whole database, apart fromthe five items already presented during the first visit (to check for a “novelty” effect). Inthe “close recommendation” condition, the algorithm was based on the characteristicsof the five items chosen at the end of the first visit. For example, if the item was amovie, another movie with the same main actor or, if no such item existed in ourdatabase, by the same director, was proposed. In the “broad recommendation”condition, the algorithm was based on the navigational data. Frequencies werecomputed and items belonging to the most frequently visited category (for instance,poetry books) were recommended.
Being translated, please wait..